sdp: refactor session/media level handling of attributes

Browse Source

The media_count meaning is horrendous. -1 means "none", a count of "0"
actually means "1". This led to various bugs in the past, so just rip it
out and use a (wmem) array from which the length can be determined.

That also means that a hard-coded limit on the media can now easily be
lifted without affecting the size of the transport_info_t structure.
(This limit, SDP_MAX_RTP_CHANNELS,  is unchanged in this patch though.)

Refactor the SDP dissector such that:

 - Media and related attributes are no longer a bunch of fixed array
   fields, but grouped in one structure. This results in the largest
   changes all over the place since "transport_info->media[n]" is now
   transformed into "media_desc->media" where "media_desc" is an element
   of the "transport_info->media_descriptions" wmem array.
 - Simplify protocol (in "m=") parsing (lots of ifs -> array + loop).
 - Remove convert_disposable_media and disposable_media_info_t, parse
   fields (media protocol from "m=", connection address from "c=", etc.)
   while parsing the SDP instead of parsing it at the end.
 - Have two distinct structures for keeping the info for the session and
   media level. Emphasize that new media descriptions are inherited from
   session level attributes (via sdp_new_media_description).
 - Delay creation of dynamic payload type information table until we
   actually create the media description. Create function
   clean_unused_media_descriptions to handle the common of freeing
   unused dynamic pt.
 - Remove SDP_IPv4/SDP_IPv6, these are replaced by checking the type
   member of the address structure.

Changes to MSRP part:

 - Move MSRP attributes to the media-level attributes.
 - Remove msrp_transport_address_set attribute, rely on the AT_NONE
   address type for detecting bad addresses.
 - Remove SDP_MSRP_IPv4 check, this never worked as the flag was never
   set. Now it relies on the address family from the host in a=path:.

Tested with these capture files with no change in PDML output nor
improvements/regressions with memleaks (as reported by ASAN):

    capture sip call wireshark 1.8.2.pcap
    NOringback.pcapng
    rtp_not_parsed_by_1_10_1.pcap
    rtsp_interleaved_coreplayer.cap
    SIP_CALL_RTP_G711.pcapng
    srtpincorrectlyselected.pcap
    tdnwifitontwifi_withnatting_clientAbhopati_03082015.pcapng

Change-Id: Ia0dbc63f8bd78cc84dad2e18174540e31b78a80d
Reviewed-on: https://code.wireshark.org/review/19072
Petri-Dish: Peter Wu <peter@lekensteyn.nl>
Tested-by: Petri Dish Buildbot <buildbot-no-reply@wireshark.org>
Reviewed-by: Anders Broman <a.broman58@gmail.com>

...

This commit is contained in:

Peter Wu 2016-12-05 00:43:02 +01:00 committed by Anders Broman

parent 104b9fe5af

commit 1d2f68b60f

1 changed files with 401 additions and 401 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Show Stats Download Patch File Download Diff File Expand all files Collapse all files

802

epan/dissectors/packet-sdp.c

View File

File diff suppressed because it is too large Load Diff